millivolt modulation of plasmonic metasurface optical response via ionic … · 2017. 6. 16. ·...

CommuniCation

1701044 (1 of 8) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

www.advmat.de

Millivolt Modulation of Plasmonic Metasurface Optical Response via Ionic Conductance

Krishnan Thyagarajan, Ruzan Sokhoyan, Leonardo Zornberg, and Harry A. Atwater*

Dr. K. Thyagarajan, Dr. R. Sokhoyan, L. Zornberg, Prof. H. A. AtwaterThomas J. Watson Laboratory of Applied PhysicsCalifornia Institute of TechnologyPasadena, CA 91125, USA E-mail: [email protected]. K. Thyagarajan, Prof. H. A. AtwaterKavli Nanoscience InstituteCalifornia Institute of TechnologyPasadena, CA 91125, USA

DOI: 10.1002/adma.201701044

liquid crystal,[8] the modulation of Fermi level in graphene via gating,[9] phase change of a material upon heating or cooling,[10] as well as field effect modula-tion of Fermi level in gallium arsenide,[11] silicon,[12] and transparent conducting oxides.[13] Notably, none of these mecha-nisms has yielded tunable metasurfaces operating in the visible wavelength range. This is partly due to challenges associ-ated with obtaining large modulation of the complex refractive index of the con-stituent materials at visible wavelengths. For example, each element of a field effect tunable metasurface is effectively a metal–insulator–semiconductor capacitor, with the metal serving as a gate and the semi-conductor functioning as a field effect

channel.[11–14] When an electrical bias is applied between the metal gate and the semiconductor, charge depletion or accu-mulation layer is formed in the semiconductor at the semicon-ductor–insulator interface. The complex dielectric permittivity of a semiconductor increases linearly with its carrier concentra-tion, resulting in field effect optical modulation. On the other hand, dielectric permittivity of a semiconductor is inversely proportional to the light frequency squared, implying larger optical modulation will be observed at smaller frequencies (longer wavelengths). So far, field effect tunable metasurfaces have only been reported for wavelengths longer than 1.5 μm. The operating wavelength of graphene-based tunable metas-urfaces has so far been limited to mid-infrared range.[15] This stems from the dielectric permittivity of graphene that is pro-portional to the square root of the carrier density in graphene and is inversely proportional to the square of the light fre-quency. Active tuning of the optical response of metasurfaces loaded with liquid crystal molecules has been previously dem-onstrated in the near-infrared wavelength range.[8,16] However, anchoring of liquid crystal molecules to the surfaces of these structures is a major drawback of this approach. This issue has prevented the demonstration of tunable metasurfaces in the visible wavelength range. Phase change materials have been previously used to demonstrate tunable metasurfaces in the mid- and near-infrared wavelength range.[10,17] Although phase change materials exhibit refractive index modulation in the vis-ible wavelength range, it is relatively modest relative to that in the near-infrared range. Here, we demonstrate tunable meta-surfaces at visible wavelengths via a modulation mechanism based on ionic conductance, and also at strikingly low (single-millivolt scale) voltages.

A plasmonic metasurface with an electrically tunable optical response that operates at strikingly low modulation voltages is experimentally demon-strated. The fabricated metasurface shows up to 30% relative change in reflectance in the visible spectral range upon application of 5 mV and 78% absolute change in reflectance upon application of 100 mV of bias. The designed metasurface consists of nanostructured silver and indium tin oxide (ITO) electrodes which are separated by 5 nm thick alumina. The millivolt-scale optical modulation is attributed to a new modulation mechanism, in which transport of silver ions through alumina dielectric leads to bias-induced nucleation and growth of silver nanoparticles in the ITO counter-electrode, altering the optical extinction response. This transport mechanism, which occurs at applied electric fields of 1 mV nm−1, provides a new approach to use of ionic transport for electrical control over light–matter interactions.

Metasurfaces

Historically, macroscopic optical components, such as lenses, curved mirrors, and polarizers have been used to control the amplitude, polarization, or the phase of light, and modify propagation over distances much larger than the wavelength. Recently, ultrathin nanostructured optical metasurfaces[1,2] whose thicknesses are comparable to or less than the wave-length have been demonstrated to achieve similar control of light by design of the metasurface phase and amplitude response.[2,3] For example, metasurface-based holograms,[4] ultrathin polarization rotators,[5] and flat focusing lens[2] have been previously demonstrated. It has also been proposed that a V-shaped antenna array can be used for on-chip transfor-mation of a single Gaussian beam into four orbital angular momentum beams.[6] However, most such metasurfaces are passive in nature, and their response is fixed once they are fabricated.

Recently, there has been considerable interest in studying dynamically tunable metasurfaces, and many different mecha-nisms have been exploited to realize active metasurfaces.[7] These include voltage-induced reorientation of molecules in a

Adv. Mater. 2017, 1701044

© 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim1701044 (2 of 8)

www.advmat.dewww.advancedsciencenews.com

Ionic conductance has been widely explored as a mechanism for electrochemical metallization-based resistive switching devices such as random access memories (RRAM).[18] A pro-totypical RRAM device consists of two electrodes and an ionic conductor sandwiched between them. The ionic conductor, which can be an oxide,[19–21] chalcogenide,[22] or halide[23] mate-rial, permits ionic transport while preventing electron trans-port. In the case of simple cation movement-based RRAMs, one electrode is oxidizable (soluble), such as silver[19,20] or copper,[21] while the counter-electrode is inert (insoluble). When a positive voltage is applied to the oxidizable electrode, its constituent metal starts to dissolve and starts depositing as a metallic filament at the opposite inert electrode (see Figure 1a). Upon increasing applied bias, the metallic filament ultimately bridges the relatively insulating ion conductor causing a large change in device resistance. Upon reversing the bias, the fila-ment starts to dissolve via ionic transport in the opposite direc-tion, and the system microstructure evolves back toward its original state. Since ionic transport causes a significant change in the resistance of the medium, the modulation mechanism is

usually referred to as resistive switching. Coupled electron-ion dynamics in resistive switching is able to realize the functional characteristic of a “memristor” (memory-resistor) which was first theoretically proposed in 1971.[24] For this reason resistive switching in RRAM devices is also referred to as memristive switching.[25]

Even though resistive switching was discovered almost 50 years ago,[26] ionic conductance-based optical modula-tion in nanophotonic systems has been demonstrated only recently.[27,28] Silver filament formation in thin amorphous sil-icon layer[27] or gold filament formation in thin silica layer[28] has been used to modulate transmittance of light through plas-monic waveguides upon application of a bias voltage. Active tuning of optical response of a plasmonic antenna attached to Ag/Al2O3/Au memristive junction has also been experimentally demonstrated.[29] Active optical response of the antenna was attributed via indirect evidence to silver filament formation in a thin Al2O3 layer. A nanoparticle-on-mirror memristive switch for modulation of scattering was experimentally demonstrated which yielded few-percent changes in scattered intensity with

Adv. Mater. 2017, 1701044

Figure 1. a) Schematic showing the migration of silver ions and formation of silver filaments upon the application of an electrical bias; gray circles in Al2O3 and ITO represent silver nanoparticles; b) transmission electron microscopy images of the planar structures showing an effective medium formed in the layers; c) schematic of the effective structure used for the simulations; d) measured reflectance spectrum from Ag/Al2O3/ITO planar heterostructure under applied bias voltages from 0 to 5 mV; and e) simulated reflectance from the planar heterostructure. The legend indicates the thickness of the effective medium layer along with corresponding fill fraction.



≈1 V applied bias,[30] and it was argued that scattered intensity from the plasmonic cavity could be used as a probe of fila-ment formation dynamics. A recent theoretical report[31] has suggested that electrochemical programming of the optical response of two adjacent Cu and Pt antennas via Cu filament formation in Cu2S layer separating the antennas. Finally, it has been experimentally demonstrated that external bias-initiated local changes in ZnO stoichiometry in Al/ZnO/Si memris-tive antennas result in observable modulation of the reflected light at mid-infrared wavelengths.[32] Our tunable metasurface instead exploits a different mechanism that operates at very low (single-millivolt scale) voltages, an ionic transport-limited nucleation process in which Ag ion migration through a thin Al2O3 dielectric induces Ag nanoparticle nucleation and growth in the indium tin oxide (ITO) counter-electrode. This mecha-nism yields a large modulation of optical transmittance and reflectance at applied electric fields of 1 mV nm−1, which is at least an order of magnitude lower than electric fields applied in the memristive optical modulation devices cited above. We would like to emphasize that in our device observed millivolt-scale optical modulation is not accompanied by large changes in the device resistance as it is the case in previously reported ionic conduction-based optical modulators.

We first investigated the reflectance modulation of Ag/Al2O3/ITO planar heterostructures when Ag is positively biased with respect to ITO, and then incorporated Ag/Al2O3/ITO het-erostructures into a metasurface to enhance the reflectance and transmittance modulation due to optical resonances supported by the structure.

The fabricated Ag/Al2O3/ITO planar heterostructures con-sist of 80 nm thick Ag, 5 nm thick Al2O3 deposited via atomic layer deposition (ALD), followed by 110 nm of ITO sputtered onto the Al2O3 layer. The carrier concentration of the fabri-cated n-type ITO film is 5 × 1020 cm−3. Transmission electron microscopy (TEM) image of the fabricated sample is shown in Figure 1b. Before being measured, the sample has been left exposed to ambient environment for one week (aged sample). Reflectance measurements for aged planar structures under applied bias are shown in Figure 1d. The measurements are performed under normal incidence with applied bias ranging from 0 to 5 mV (with 1 mV steps). Upon application of a voltage, the reflectance decreases at longer wavelengths and increases at shorter wavelengths, with a crossover region at around 650 nm. The measured reflectance is much lower than that obtained from the simulations performed with complex refractive indices for constituent layers obtained via spectro-scopic ellipsometry measurements performed for each con-stituent layer separately (see Part S2, Supporting Information). We hypothesize that this could be due to migration of silver ions into alumina and possibly also into ITO. We assume that the effective refractive indices of composite media consisting of intermixtures of alumina and silver and intermixtures of ITO and silver can be described by using an effective medium model in the Bruggeman approximation.[33] As a first step, we assume that Ag ions penetrate only into the 5 nm thick Al2O3 layer and calculate the reflectance from Ag/Al2O3/ITO planar stack. However, the experimentally observed reflectance changes cannot be reproduced in simulations that assume Ag penetrates only into the thin Al2O3 layer. However, simulations

that also assume silver ion penetration into the 110 nm thick ITO layer are able to match the experimentally observed reflec-tance changes. Based on electron microscopy observations (see below), we modeled the ITO layer as being subdivided into two layers: (i) a homogeneous layer of ITO located at the top inter-face of Al2O3 and (ii) an Ag nanoparticle-containing ITO layer which is treated as an effective medium, located between the homogeneous ITO layer and the free surface (see schematic shown in Figure 1c). Using the transfer matrix method, we cal-culate the reflectance from the multilayer stack and fit it to the experimentally measured curves. The thickness of the effective medium Ag/ITO layer and fill fraction of Ag in this layer are used as fitting parameters. The resulting simulations match the experimental data well (see Figure 1e). Each curve shown in Figure 1e corresponds to a distinct Ag nanoparticle fill frac-tion in the ITO layer and distinct thickness of the effective medium layer. Comparison of the fitted theoretical and experi-mental curves indicates that, when an electrical bias is applied, the effective medium layer shrinks in thickness from 64 to 47 nm, while the fill fraction of Ag nanoparticles in the effec-tive medium layer increases from 11% to 20%. The legend in Figure 1e indicates the thickness of the effective medium layer along with the Ag nanoparticle fill fraction. Transmission elec-tron microscopy images of the planar heterostructure indicate intermixing of Ag and ITO (Figure 1b).

Finally, to verify our hypothesis that the observed optical modulation is caused by migration of Ag ions, we obtained electron microscopy images of aged planar structures fabri-cated on Si substrates. In these samples, an additional 10 nm thick Al2O3 layer has been deposited on Si to electrically isolate the Ag layer and the underlying substrate. We performed elec-tron microscopy imaging on two different aged planar struc-tures: (i) first sample has never been electrically modulated and (ii) second sample was subjected to over 100 modulation cycles with applied electrical biases ranging from 0 to 5 mV. The electron microscopy images gathered from an unmodu-lated sample are shown in Figure 2a,b. One can clearly see well-defined material layers with no material intermixing. On the other hand, the images taken from the repeatedly modu-lated sample are given in Figure 2c–g. Figure 2c shows that there are 5–12 nm diameter particles in ITO, which increase in concentration and size near the ITO/air interface (see encircled region). Local energy-dispersive X-ray spectroscopy (EDS) meas-urements indicate an Ag fraction of up to 5.2% in those regions of the ITO. Figure 2d,f shows electron microscopy images for samples that have been repeatedly modulated with bias voltages from 0 to 5 mV. The high-magnification images in Figure 2d–g reveal a polycrystalline material which appears to be Ag inter-mixed with Al2O3. Figure 2d (2f) is identical to Figure 2e (2g) but the spatial regions of intermixture of the Al2O3 and Ag are highlighted with dashed outlines. These images confirm that the observed optical modulation is indeed due to migration of Ag ions under applied bias (for additional TEM images and EDS data see Parts S3–S6, Supporting Information). Previously, conducting Ag filament formation in Ag/Al2O3/ITO memris-tive structures has been imaged via dual beam time-of-flight secondary ions mass spectrometry.[34] It has been observed that under applied bias Ag ions may penetrate into ITO for hun-dreds of nanometers, which is consistent with our results. In

Adv. Mater. 2017, 1701044



our case, no electroforming step is necessary to trigger the ion migration process. Other work analyzing the resistive switching in Pt/TiO2/Pt structures also indicated that no electroforming process is necessary to trigger ionic transport when the gate dielectric is thinner than 6 nm.[35] On the other hand, resistive switching under millivolt bias with no prior electroforming step has been recently observed in a Pt/TiO2/TiN structure.[36] The thickness of TiO2 in this structure was 70 nm. This observation suggests that the presence of the TiN top electrode is crucial for elimination of electroforming and ultralow bias operation. Similarly, our results suggest that the ITO electrode is essen-tial for observation of optical modulation under millivolt bias. By contrast, an applied bias voltage on the order of volts was required to enable optical modulation in Ag/Al2O3/Au crossbar antenna devices[29]—which do not contain ITO—even though the Al2O3 layer was only 5 nm thick. For Pt/TiO2/Pt structures, it is known that resistive switching occurs due to movement of oxygen vacancies. In case of the Ag/Al2O3/ITO heterostructures studied here, both migration Ag ions and oxygen vacancies are likely contributors to transport and device operation. Indeed, it has been previously demonstrated that ITO layers incorporated in a memristive device can absorb or provide oxygen vacancies under an applied bias.[37] Furthermore, one may imagine that the change in ionic concentration at the ITO/Al2O3 interface may be dictating the observed optical modulation. However, this conjecture can be disproved by looking at the TEM images as shown in Figure 2b,c. It can be seen that the silver nano-particles actually penetrate into the ITO and that there are no particles located at the ITO and ITO/Al2O3 interface. In addi-tion to this, simulations on the planar interfaces show indeed

that the observed modulation can be explained by incorporating silver nanoparticles into an effective medium like approxima-tion, as seen in Figure 1d,e. This thus leads us to believe that the observed phenomenon is in fact of a new origin.

Next, we incorporate Ag/Al2O3/ITO tunable heterostruc-tures into a metasurface. Optical resonances supported by the metasurface enhance light–matter interactions, resulting in larger tunability of the reflectance and transmittance when a bias voltage is applied. The fabricated tunable metasurface con-sists of quartz substrate, 1 nm Cr adhesion layer, followed by 80 nm thick Ag in which inverse dolmen structures have been milled using focused ion beam lithography (see Figure 3a–c). The inverse dolmen structure consists of two lower parallel rods and a top rod. The lower parallel rods are 150 nm long and 100 nm wide, and their mutual spacing is 30 nm. Both of these are 30 nm away from a 230 nm long and 100 nm wide top rod. A scanning electron microscope (SEM) image of an array of fabricated inverse dolmen structures is shown in Figure 3c. Next, we conformally deposit 5 nm Al2O3 films via atomic layer deposition, and 110 nm of ITO via RF magnetron sputtering. A schematic of the light–matter interaction in such structures is shown in Figure 3a, with the inset depicting different layers. The total area of the fabricated inverse dolmen structures is 200 μm × 200 μm. Figure 3b shows a schematic of the cross section of the structure.

When milling inverse dolmen structures into a silver film, a thin silver layer was left at the bottom of the device (between chromium and bottom Al2O3 layers). Two sets of inverse dolmen samples were measured: (i) first set of samples was measured immediately upon fabrication (fresh samples) and (ii) second

Adv. Mater. 2017, 1701044

Figure 2. TEM images of the cross section of the planar structure. a) A sample that has never been electrically modulated; b) a compressed image of (a); c) a sample that has been subject to >103 cycles of electrical bias modulation thereby showing the presence of silver up even inside the ITO layer (dark spots encircled in a back dashed rectangle); scale bars in (b) and (c) correspond to 80 nm; the compressed image (b) is depicted for facilitating the comparison with (c); d–g) magnified image of the interface between Ag/Al2O3/ITO of the heavily electrically modulated sample shown in (c). The scale bar in (d)–(g) is 5 nm. d) Intermixing of Al2O3 and silver in the dielectric layer and f) formation of silver inclusions inside the Al2O3 layer. Parts (e) and (g) are identical to (d) and (f), except that the regions of interest have been highlighted by white dashed lines.



set of samples had been left exposed to ambient environment for a week before being measured (aged samples). Interestingly, we observe that reflectance and transmittance spectra of the fresh samples differ from the spectra obtained from the aged samples. Moreover, aged samples show much greater optical reflectance and transmittance modulation than the fresh ones. We believe that the origin of the “aged sample effect” can be attributed to the contamination of the samples by chlorine and sulfur, both of which are known to assist in the mobility of silver ions and ease in formation of ions.[38] The results of the optical measurements performed on fresh samples are shown

in Figure 3d,f,h. First, measurements were taken at zero bias and subsequently silver electrode was positively biased with respect to ITO electrode. Upon the application of a 5 mV bias (in five steps, leading to each incremental graph representing a bias of 1 mV), the reflectance of fresh samples exhibits a broad-band relative increase by up to 9% that corresponds to the abso-lute reflectance increase by nearly 6%. The relative reflectance increase is defined by 100 × [R(5 mV) − R(0 mV)]/R(0 mV), where R(0 mV) and R(5 mV) stand for reflectance values at no applied bias and 5 mV of bias, respectively. As one can see from Figure 3f, the measured transmittance exhibits a broadband

Adv. Mater. 2017, 1701044

Figure 3. a) Schematic of the envisioned light–matter interaction in the metasurface with dimensions mentioned; b) cross section showing the dif-ferent layers in the structures; c) SEM images of the inverse dolmen structures showing an array and individual configuration; optical measurements of d,e) reflectance, f,g) transmittance, and h,i) absorptance spectra at normal incidence for fresh and aged dolmen samples, respectively. The dashed lines show simulation results at no applied bias, while the solid lines represent the experimental results under applied bias from 0 to 5 mV. Applied bias varies from 0 to 5 mV with incremental steps of 0.5 mV.



relative decrease by up to 35% that corresponds to the absolute transmittance decrease by around 6%. Interestingly, upon bias voltage application, dips and peaks observed in reflectance and transmittance do not change their spectral position. Results of the full wave electromagnetic simulations under no applied bias are indicated by dashed lines (for further details see Part S7, Supporting Information). As one can see, in the case of fresh samples, the simulated and experimental curves match reasonably well. Next, we performed optical measurements on aged dolmen samples. Reflectance and transmittance spectra measured under an applied bias are shown in Figure 3e,g,i. As one can see, the aged samples show a much larger reflectance modulation than fresh ones: we observe relative reflectance increase by 30% (15% absolute increase) and relative transmit-tance decrease by 33% (5% absolute decrease) upon application of a 5 mV bias voltage. Spectral characteristics of the aged sam-ples at zero bias are also different from those obtained from fresh samples: aged samples show lower reflectance at shorter wavelengths and higher transmittance at longer wavelengths. For increasing bias voltages, the reflectance increases until it reaches an absolute maximum saturation value of 90% upon the application of 6 mV (see Part S8, Supporting Information). Thereafter, there is a steep reduction in the reflectance, tailing down at 12% beyond which there is a dramatic increase in the leakage current. This gives a strikingly large dynamic range of absolute reflectance change by ΔR = 78% with the applica-tion of just 100 mV (for additional optical measurements, see Parts S9–S11, Supporting Information). We have also per-formed cyclic voltammetry measurements on aged dolmen samples. We observe a pinched hysteresis loop in a plot of cur-rent versus voltage that is a signature of a memristive system[39] (see Part S13, Supporting Information). One of the important figures of merit of RRAM devices is defined by the ratio of the device resistances in the on and off states.[18] However, within the context of ionic conductance-based optical modulators the change of the device resistance is not an important metric

any more since large optical modulation can still be observed without considerable resistance changes. For example, under millivolt electrical bias our device shows significant optical modulation without large changes in resistance. Hence, in this regime, the distribution of silver ions in our device can be sche-matically described by the middle panel of Figure 1a.

Finally, we identify the response time of the fabricated devices. To characterize the frequency response of the aged planar structures as well as aged dolmen metasurfaces, a 5 mV bias, with frequencies ranging from 100 to 600 Hz, was applied and a high-speed Si detector was used to detect the reflectance from the structures. The measurements have been done at the wavelength of λ = 700 nm. The reflectance values plotted in Figure 4a,b are normalized to the values of reflectance change obtained at 100 Hz modulation frequency. As modulation frequency increases, the amplitude of the modulated signal decreases. At modulation frequency of 600 Hz, the recorded normalized signal is almost flat implying that the silver ions are unable to get fully rearranged upon electrical modulation. Thus, frequency response of planar structures and inverse dolmen metasurfaces is very similar. We can make an estimate of the silver nanoparticle formation dynamics by looking at the trans-mission electron microscopy images shown in Figure 2b,c and referring to Figure 4. From Figure 2c we note the radius of the bigger silver nanoparticles is 10–12 nm. For a density of face-centered cubic (FCC) silver of 10.5 g cm−3, the mass of a single silver nanoparticle is of order of 4.4 × 10−20 kg. Considering the total sample area and the thickness of ITO, the volume into which silver ions can penetrate is roughly 1.38 × 10−14 m3. From Figure 1d,e, we can estimate the volume of silver penetrated into ITO at the two extreme biases of 0 V and 5 mV, yielding fill fractions of 11% and 20%, and Ag volumes of 1.52 × 10−15 and 2.76 × 10−15 m3, respectively. In turn, we obtain a number of silver atoms moving into or out of the ITO of 3.63 × 108 and 6.6 × 108, respectively, at the two fill fractions. Since silver ions have unit charge, we can estimate the amount of charge

Adv. Mater. 2017, 1701044

Figure 4. Reflectance signal as a function of time for different modulation frequencies for a) an aged dolmen metasurface and b) an aged planar structure. In cases of both planar structures and metasurfaces reflectance values are normalized to the values of the reflectance change obtained at 100 Hz modulation frequency.



passed into or out of the ITO volume at 0 V and 5 mV biases. For the 100 Hz signal modulation at 5 mV in Figure 4, we obtain a charge movement of 2.75 × 10−11 C in 1 ms. This cor-responds to a current of 27.5 nA, which compares to a value of 15 nA obtained in our I–V plots in Figure S16 (Supporting Information). This quick calculation supports our hypothesis that the AC modulation current is sufficient to explain the Ag ion transport needed to nucleate the nanoparticles that give rise to optical modulation. The moderate modulation frequency reported here is partially due to large device area (1 mm × 1 mm) that results in large RC constant. When studying Ag ion migration in small-area Ag/ZnO:Mn/Pt device, switching frequencies on the order of 100 MHz have been observed.[19] Since power consumption of an electronic device scales as P = CV2f, where C is the capacitance of the device, V is the applied bias, and f is the modulation frequency, shrinking area of our device could pave a route toward ultralow-power optical modulators.

In this Communication, we have experimentally demon-strated an ultralow voltage optical modulator exhibiting abso-lute reflectance modulation by up to 78% with a bias of 100 mV (and absolute reflectance increase by 15% with a bias of just 5 mV). Our system exploits a new mechanism of ionic con-ductance-mediated nucleation and growth of Ag nanoparticles to modulate the optical reflectance and transmittance. Since complex dielectric permittivity values of ITO and Ag are very different in a broad wavelength range, one may expect that the reported modulation mechanism may yield active optical devices for wavelengths extending beyond visible (see Part S14, Supporting Information). The results represent an advance over previous work[27–30] by revealing a new mechanism that operates at dramatically lower electric fields, yielding a reduc-tion of the applied bias voltages for optical reflectance modula-tion by three orders of magnitude.

Our work could be used for design of future attojoule-energy scale optical modulators, highly sensitive optical readout of sur-face phenomena, optical display technology, and nonvolatile optical memories.

Experimental SectionFabrication and Characterization of Planar Ag/Al2O3/ITO

Heterostructures: To fabricate Ag/Al2O3/ITO planar heterostructures, a 1 nm Cr adhesion layer and 80 nm of Ag were evaporated, using electron beam deposition, on either Si or SiO2 substrates. Next, 5 nm of Al2O3 via atomic layer deposition (ALD) and 110 nm of ITO via RF magnetron sputtering (O2 flow rate of 0.5 sccm) were deposited. The samples fabricated on SiO2 substrate were used to perform optical measurements. The samples fabricated on Si substrates contain a 10 nm thick Al2O3 buffer layer that electrically insulates Ag/Al2O3/ITO planar stack from Si (Figure 1b). The samples on Si substrate are used for TEM imaging. Materials characterization of the ITO revealed the carrier concentration and permittivity, using Van der Pauw Hall measurements and spectroscopic ellipsometry, respectively. The resulting film was n-type with a carrier concentration of 5 × 1020 cm−3. To perform Hall measurements, 110 nm of ITO was deposited on SiO2 substrates, while for ellipsometry measurements, 110 nm of ITO was deposited on Si substrates. The 5 nm layer of ALD-deposited Al2O3 was repeatedly tested for the leakage current and it was seen that most of the samples were displaying very good insulation properties. For a device with area of 1 mm × 1 mm a 5–10 nA leakage current was observed. The planar

samples on SiO2 substrate enable one to perform both reflectance and transmittance measurements. However, due to the large Ag thickness, the measured transmittance was negligibly small.

Optical Measurements: Optical measurements under applied bias were undertaken at normal incidence. A microscope objective 5× (Olympus, NA 0.6) was used to focus incident light from a supercontinuum laser light source to a spot of size roughly 20 μm, and the signal was measured using a Si detector. Measurements were taken for a wavelength range from 500 to 800 nm. For the bias-dependent measurements, a voltage was applied using a voltage source. The bias was applied between the top layer of ITO and the bottom layer of silver at locations far from the measured area.

Transmission Electron Microscopy: Transmission electron microscopy/EDS analysis was performed with FEI TF30ST transmission electron microscope at an accelerating voltage of 300 kV.

Fabrication of the Tunable Metasurface: To fabricate the tunable metasurface, a 1 nm chromium adhesion layer and 80 nm of silver were evaporated, using electron beam deposition, on a quartz substrate. Focused ion beam lithography was carried out on the deposited silver film to create a metasurface of inverse dolmen structures (see Figure 3a–c). Next, 5 nm Al2O3 films via atomic layer deposition and 110 nm of ITO via RF magnetron sputtering (O2 flow rate of 0.5 sccm) were deposited. When milling inverse dolmen structures into a silver film, a thin silver layer was left at the bottom of the device (between chromium and bottom Al2O3 layers). All the deposition steps were undertaken with appropriate face masks made of stainless steel, to permit an easy bias application configuration in the final device. A transmission electron microscopy image of an identical planar heterostructure (Cr/Ag/Al2O3/ITO) deposited on a silicon substrate showed a uniform layer of roughly 1.2 nm of chromium as the adhesion layer (see Figure S3, Supporting Information) and hence no disconnected islands of chromium were created.

SEM and EDS: Dual beam focused ion beam with X-ray microanalysis FEI Nova 200 Nanolab was used to gather SEM images. The same tool was used to perform EDS characterization of the dolmen samples (see Parts S5 and S6, Supporting Information). Besides detecting the presence of silver, oxygen, indium, and tin, a finite sulfur and chlorine signal was present throughout, wherever silver was present, for the aged sample. The presence of sulfur and chlorine is attributed to silver tarnishing upon air exposure during the fabrication process and during subsequent ambient atmosphere exposure.

Calculation of the Optical Response of the Structures: The reflectance from Ag/Al2O3/ITO planar heterostructures was calculated by using transfer matrix method. Effective dielectric permittivity of the composite media consisting of intermixtures of Al2O3 and Ag and intermixtures of ITO and Ag is described by using an effective medium model in the Bruggeman approximation. Full-wave simulations were performed to calculate the optical response of the designed inverse dolmen metasurface (Part S7, Supporting Information). For full-wave simulations finite-difference time-domain method was used.

Supporting InformationSupporting Information is available from the Wiley Online Library or from the author.

AcknowledgementsThis work was supported by the Swiss National Science Foundation (Grant No. 151853), Hybrid Nanophotonics Multidisciplinary University Research Initiative Grant (Air Force Office of Scientific Research, FA9550-12-1-0024), and the Samsung Electronics Metaphotonics Cluster. The conducting oxide material synthesis design and characterization was supported by the U.S. Department of Energy (DOE), Office of Science Grant (DE-FG02-07ER46405). Used facilities were supported by the Kavli

Adv. Mater. 2017, 1701044



Adv. Mater. 2017, 1701044

Nanoscience Institute (KNI) and Joint Center for Artificial Photosynthesis (JCAP) at Caltech. The authors would like to thank Artur R. Davoyan, Matthew Sullivan Hunt, and Barry Baker for useful discussions; Carol Garland for help with the TEM imaging; and Jonathan Grandidier for his help with the transfer matrix code.

Conflict of InterestThe authors declare no conflict of interest.

Keywordsfilament formation, indium tin oxide (ITO), ionic transport, memristors, tunable metasurfaces

Received: February 21, 2017Revised: April 28, 2017

Published online:

[1] J. Wang, J. Du, Appl. Sci. 2016, 6, 239.[2] A. Arbabi, Y. Horie, M. Bagheri, A. Faraon, Nat. Nanotechnol. 2015,

10, 937.[3] a) D. Veksler, E. Maguid, N. Shitrit, D. Ozeri, V. Kleiner, E. Hasman,

ACS Photonics 2015, 2, 661; b) A. Pors, M. G. Nielsen, R. L. Eriksen, S. I. Bozhevolnyi, Nano Lett. 2013, 13, 829; c) N. F. Yu, F. Capasso, Nat. Mater. 2014, 13, 139.

[4] a) W. T. Chen, K.-Y. Yang, C.-M. Wang, Y.-W. Huang, G. Sun, I.-D. Chiang, C. Y. Liao, W.-L. Hsu, H. T. Lin, S. Sun, L. Zhou, A. Q. Liu, D. P. Tsai, Nano Lett. 2014, 14, 225; b) G. Zheng, H. Muhlenbernd, M. Kenney, G. Li, T. Zentgraf, S. Zhang, Nat. Nanotechnol. 2015, 10, 308; c) Y.-W. Huang, W. T. Chen, W.-Y. Tsai, P. C. Wu, C.-M. Wang, G. Sun, D. P. Tsai, Nano Lett. 2015, 15, 3122; d) X. J. Ni, A. V. Kildishev, V. M. Shalaev, Nat. Commun. 2013, 4, 2807.

[5] C. Pfeiffer, C. Zhang, V. Ray, L. J. Guo, A. Grbic, Optica 2016, 3, 427.[6] J. Du, J. Wang, Sci. Rep. 2015, 5, 9662.[7] M. Ferrera, N. Kinsey, A. Shaltout, C. DeVault, V. Shalaev,

A. Boltasseva, J. Opt. Soc. Am. B 2017, 34, 95.[8] O. Buchnev, N. Podoliak, M. Kaczmarek, N. I. Zheludev,

V. A. Fedotov, Adv. Opt. Mater. 2015, 3, 674.[9] V. W. Brar, M. S. Jang, M. Sherrott, J. J. Lopez, H. A. Atwater, Nano

Lett. 2013, 13, 2541.[10] M. J. Dicken, K. Aydin, I. M. Pryce, L. A. Sweatlock, E. M. Boyd,

S. Walavalkar, J. Ma, H. A. Atwater, Opt. Express 2009, 17, 18330.[11] Y. C. Jun, J. Reno, T. Ribaudo, E. Shaner, J. J. Greffet, S. Vassant,

F. Marquier, M. Sinclair, I. Brener, Nano Lett. 2013, 13, 5391.[12] A. Olivieri, C. Chen, S. Hassan, E. Lisicka-Skrzek, R. N. Tait,

P. Berini, Nano Lett. 2015, 15, 2304.[13] Y. W. Huang, H. W. H. Lee, R. Sokhoyan, R. A. Pala, K. Thyagarajan,

S. Han, D. P. Tsai, H. A. Atwater, Nano Lett. 2016, 16, 5319.[14] J. Park, J. H. Kang, X. G. Liu, M. L. Brongersma, Sci. Rep. 2015, 5,

15754.[15] a) M. C. Sherrott, P. W. C. Hon, K. T. Fountaine, J. C. Garcia,

S. M. Ponti, V. W. Brar, L. A. Sweatlock, H. A. Atwater, Nano Lett. 2017, 17 (5), 3027-3034; b) N. Dabidian, I. Kholmanov, A. B. Khanikaev, K. Tatar, S. Trendafilov, S. H. Mousavi,

C. Magnuson, R. S. Ruoff, G. Shvets, ACS Photonics 2015, 2, 216; c) Y. Yao, M. A. Kats, P. Genevet, N. F. Yu, Y. Song, J. Kong, F. Capasso, Nano Lett. 2013, 13, 1257.

[16] a) J. Sautter, I. Staude, M. Decker, E. Rusak, D. N. Neshev, I. Brener, Y. S. Kivshar, ACS Nano 2015, 9, 4308; b) M. Decker, C. Kremers, A. Minovich, I. Staude, A. E. Miroshnichenko, D. Chigrin, D. N. Neshev, C. Jagadish, Y. S. Kivshar, Opt. Express 2013, 21, 8879.

[17] a) Y. G. Chen, T. S. Kao, B. Ng, X. Li, X. G. Luo, B. Luk’yanchuk, S. A. Maier, M. H. Hong, Opt. Express 2013, 21, 13691; b) B. Gholipour, J. F. Zhang, K. F. MacDonald, D. W. Hewak, N. I. Zheludev, Adv. Mater. 2013, 25, 3050.

[18] F. Pan, S. Gao, C. Chen, C. Song, F. Zeng, Mater. Sci. Eng., R 2014, 83, 1.

[19] Y. C. Yang, F. Pan, Q. Liu, M. Liu, F. Zeng, Nano Lett. 2009, 9, 1636.[20] Y. C. Yang, P. Gao, S. Gaba, T. Chang, X. Q. Pan, W. Lu, Nat.

Commun. 2012, 3, 732.[21] S. Yazdanparast, J. A. Koza, J. A. Switzer, Chem. Mater. 2015, 27,

5974.[22] F. Zhuge, K. Li, B. Fu, H. L. Zhang, J. Li, H. Chen, L. Y. Liang,

J. H. Gao, H. T. Cao, Z. M. Liu, H. Luo, AIP Adv. 2015, 5, 057125.[23] X. F. Liang, Y. Chen, L. Shi, J. Lin, J. Yin, Z. G. Liu, J. Phys. D: Appl.

Phys. 2007, 40, 4767.[24] L. O. Chua, IEEE Trans. Circuit Theory 1971, ct-18, 507.[25] J. J. Yang, M. D. Pickett, X. M. Li, D. A. A. Ohlberg, D. R. Stewart,

R. S. Williams, Nat. Nanotechnol. 2008, 3, 429.[26] T. W. Hickmott, J. Appl. Phys. 1962, 33, 2669.[27] a) A. Emboras, I. Goykhman, B. Desiatov, N. Mazurski, L. Stern,

J. Shappir, U. Levy, Nano Lett. 2013, 13, 6151; b) A. Emboras, J. Niegemann, P. Ma, C. Haffner, A. Pedersen, M. Luisier, C. Hafner, T. Schimmel, J. Leuthold, Nano Lett. 2016, 16, 709.

[28] C. Hoessbacher, Y. Fedoryshyn, A. Emboras, A. Melikyan, M. Kohl, D. Hillerkuss, C. Hafner, J. Leuthold, Optica 2014, 1, 198.

[29] D. T. Schoen, A. L. Holsteen, M. L. Brongersma, Nat. Commun. 2016, 7, 12162.

[30] G. Di Martino, S. Tappertzhofen, S. Hofmann, J. Baumberg, Small 2016, 12, 1334.

[31] S. Dong, K. Zhan, Z. P. Yu, J. A. Fan, ACS Nano 2016, 10, 6716.[32] E. Battal, A. Ozcan, A. K. Okyay, Adv. Opt. Mater. 2014, 2, 1149.[33] G. A. Niklasson, C. G. Granqvist, O. Hunderi, Appl. Opt. 1981, 20,

26.[34] Y. Busby, S. Nau, S. Sax, E. J. W. List-Kratochvil, J. Novak,

R. Banerjee, F. Schreiber, J. J. Pireaux, J. Appl. Phys. 2015, 118, 075501.

[35] J. J. Yang, F. Miao, M. D. Pickett, D. A. A. Ohlberg, D. R. Stewart, C. N. Lau, R. S. Williams, Nanotechnology 2009, 20, 215201.

[36] A. Al-Haddad, C. L. Wang, H. Y. Qi, F. Grote, L. Y. Wen, J. Bernhard, R. Vellacheri, S. Tarish, G. Nabi, U. Kaiser, Y. Lei, ACS Appl. Mater. Interfaces 2016, 8, 23348.

[37] a) E. O. Filatova, A. P. Baraban, A. S. Konashuk, M. A. Konyushenko, A. A. Selivanov, A. A. Sokolov, F. Schaefers, V. E. Drozd, New J. Phys. 2014, 16, 113014; b) T.-C. Chang, C.-H. Pan, T.-M. Tsai, K.-C. Chang, T.-J. Chu, C.-C. Shih, C.-Y. Lin, P.-H. Chen, H. Wu, N. Deng, H. Qian, S. M. Sze, IEEE Trans. Electron. Devices 2016, 63, 4737.

[38] a) H. K. Hardy, Prog. Met. Phys. 1956, 6, 45; b) G. W. Luckey, Discuss. Faraday Soc. 1959, 28, 113; c) S. P. Xie, X. C. Zhang, D. Xiao, M. C. Paau, J. Huang, M. M. F. Choi, J. Phys. Chem. C 2011, 115, 9943; d) Y. X. Zhang, S. D. Sun, X. Z. Zhang, L. L. Tang, X. P. Song, Z. M. Yang, Phys. Chem. Chem. Phys. 2014, 16, 18918.

[39] a) R. Tetzlaff, Memristors and Memristive Systems, Springer, New York 2014; b) Y. C. Yang, W. Lu, Nanoscale 2013, 5, 10076.

millivolt modulation of plasmonic metasurface optical response via ionic … · 2017. 6. 16. ·...

Documents